A subspace reverse-correlation technique for the study of visual neurons.

A new discrete-time reverse-correlation scheme for the study of visual neurons is proposed. The visual stimulus is generated by drawing with uniform probability, at each refresh time, an image from a finite set S of orthonormal images. We show that if the neuron can be modeled as a spatiotemporal linear filter followed by a static nonlinearity, the cross-correlation between the input image sequence and the cell's spike train output gives the projection of the receptive field onto the subspace spanned by S. The technique has been applied to the analysis of simple cells in the primary visual cortex of cats and macaque monkeys. Experimental results are presented where S spans a subspace of spatially low-pass signals. Advantages of the proposed scheme over standard white-noise techniques include improved signal to noise ratios, increased spatial resolution, and the possibility to restrict the study to particular subspaces of interest.